Skip to content
Interactive in-browser attention visualizer tool for recurrent networks
HTML
Branch: master
Clone or download
Latest commit ecf8add Apr 30, 2017
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
external first commit Jan 30, 2017
README.md change readme Apr 30, 2017
attn_vis_data.json fix example data Apr 30, 2017
index.html change to work with release of tensorflow code Apr 30, 2017

README.md

Attention Visualizer

This is a tool to visualize the distribution of attention in a text-based sequence-to-sequence task such as summarization. As you hover your mouse over the decoded words, the tool shows a heatmap of attention over the source words. A demo can be seen here (scroll down to "Example Output" section).

Additionally, for pointer-generator networks such as that described in this paper, the tool displays the generation probability of each decoded word. This tool was designed to work with the Tensorflow code for the paper.

To run

To run the visualizer, run

python -m SimpleHTTPServer

from this directory then navigate to http://localhost:8000/ in browser. The visualizer will show some example data.

To use your own data

To visualize your own data, you need to replace attn_vis_data.json with a similar file, either produced by this Tensorflow code, or by your own model. In particular attn_vis_data.json should contain the following fields:

  • article_lst: the article (or source text) as a list of words
  • decoded_lst: the decoded (i.e. machine-generated) summary as a list of words
  • abstract_str: the reference summary as a single string
  • attn_dists: a list same length as decoded_lst, containing lists of length "attention length", containing probabilities. Note attention length must be <= length of article_lst. e.g. your article may have 500 words but you only fed the first 200 words into the model, thus attention length is 200. In this case the visualizer will mark the truncation point in the article.
  • p_gens: a list same length as decoded_lst, containing the generation probabilities.

WARNING: Make sure that none of the strings in article_lst, decoded_lst, or abstract_str contain <angled brackets>. These will interfere with the HTML and can result in text not being displayed.

You can’t perform that action at this time.